train_results.json

{
    "entropy": 0.03382568359375,
    "epoch": 2.0,
    "mean_token_accuracy": 0.9902507376670837,
    "num_tokens": 4840215.0,
    "total_flos": 6582062678016.0,
    "train_loss": 0.05680735134297893,
    "train_runtime": 734.9157,
    "train_samples": 2631,
    "train_samples_per_second": 7.16,
    "train_steps_per_second": 0.114
}